{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HWJORDD67BIQPHTJVFISXJOJCP","short_pith_number":"pith:HWJORDD6","schema_version":"1.0","canonical_sha256":"3d92e88c7ef851079e69a9512ba5c913e0689db1dcb03af30eb2d8d33a50abc0","source":{"kind":"arxiv","id":"2606.19659","version":1},"attestation_state":"computed","paper":{"title":"SAGE-OPD: Selective Agent-Guided Intervention for Multi-Turn On-Policy Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bo Peng, Jiayi Liu, Lizhu Zhang, Mingyi Wang, Xiangjun Fan, Yifan Wu, Yuhang Zhou, Zhuokai Zhao","submitted_at":"2026-06-17T23:58:14Z","abstract_excerpt":"On-policy distillation (OPD) improves student models by training them on trajectories induced by their own policy, making it a promising approach for mitigating exposure bias in agent training. However, most OPD studies focus on single-turn settings, while realistic LLM agents interact with environments over multiple turns. In this regime, early errors can alter future observations and compound across the trajectory, and standard dense token-level OPD becomes brittle, as it may over-penalize semantically valid alternatives, reinforce local degeneracies such as repeated actions, and propagate u"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2606.19659","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-17T23:58:14Z","cross_cats_sorted":[],"title_canon_sha256":"6917545833bb91fa11cb95a9c0ea18efee55dcd2f74685eef004b2012f0503da","abstract_canon_sha256":"f70112e539f2d2c9aa6f5d11eb7679351f8353168dab77fe831eeafde92a60ec"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:12:31.519071Z","signature_b64":"OORFM2G/sqXVtOelqLWMckYjXtXsCFsfGhgDYMBvh1ahltqILtSZO7r3vw/o3/KiD24rnOBCzcUokF19L5HgDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3d92e88c7ef851079e69a9512ba5c913e0689db1dcb03af30eb2d8d33a50abc0","last_reissued_at":"2026-06-19T16:12:31.518720Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:12:31.518720Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SAGE-OPD: Selective Agent-Guided Intervention for Multi-Turn On-Policy Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bo Peng, Jiayi Liu, Lizhu Zhang, Mingyi Wang, Xiangjun Fan, Yifan Wu, Yuhang Zhou, Zhuokai Zhao","submitted_at":"2026-06-17T23:58:14Z","abstract_excerpt":"On-policy distillation (OPD) improves student models by training them on trajectories induced by their own policy, making it a promising approach for mitigating exposure bias in agent training. However, most OPD studies focus on single-turn settings, while realistic LLM agents interact with environments over multiple turns. In this regime, early errors can alter future observations and compound across the trajectory, and standard dense token-level OPD becomes brittle, as it may over-penalize semantically valid alternatives, reinforce local degeneracies such as repeated actions, and propagate u"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.19659","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2606.19659/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2606.19659","created_at":"2026-06-19T16:12:31.518779+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.19659v1","created_at":"2026-06-19T16:12:31.518779+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.19659","created_at":"2026-06-19T16:12:31.518779+00:00"},{"alias_kind":"pith_short_12","alias_value":"HWJORDD67BIQ","created_at":"2026-06-19T16:12:31.518779+00:00"},{"alias_kind":"pith_short_16","alias_value":"HWJORDD67BIQPHTJ","created_at":"2026-06-19T16:12:31.518779+00:00"},{"alias_kind":"pith_short_8","alias_value":"HWJORDD6","created_at":"2026-06-19T16:12:31.518779+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HWJORDD67BIQPHTJVFISXJOJCP","json":"https://pith.science/pith/HWJORDD67BIQPHTJVFISXJOJCP.json","graph_json":"https://pith.science/api/pith-number/HWJORDD67BIQPHTJVFISXJOJCP/graph.json","events_json":"https://pith.science/api/pith-number/HWJORDD67BIQPHTJVFISXJOJCP/events.json","paper":"https://pith.science/paper/HWJORDD6"},"agent_actions":{"view_html":"https://pith.science/pith/HWJORDD67BIQPHTJVFISXJOJCP","download_json":"https://pith.science/pith/HWJORDD67BIQPHTJVFISXJOJCP.json","view_paper":"https://pith.science/paper/HWJORDD6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.19659&json=true","fetch_graph":"https://pith.science/api/pith-number/HWJORDD67BIQPHTJVFISXJOJCP/graph.json","fetch_events":"https://pith.science/api/pith-number/HWJORDD67BIQPHTJVFISXJOJCP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HWJORDD67BIQPHTJVFISXJOJCP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HWJORDD67BIQPHTJVFISXJOJCP/action/storage_attestation","attest_author":"https://pith.science/pith/HWJORDD67BIQPHTJVFISXJOJCP/action/author_attestation","sign_citation":"https://pith.science/pith/HWJORDD67BIQPHTJVFISXJOJCP/action/citation_signature","submit_replication":"https://pith.science/pith/HWJORDD67BIQPHTJVFISXJOJCP/action/replication_record"}},"created_at":"2026-06-19T16:12:31.518779+00:00","updated_at":"2026-06-19T16:12:31.518779+00:00"}